Relaxation kinetics of glutamate dehydrogenase self-association by

Relaxation kinetics of glutamate dehydrogenase self-association by pressure perturbation. Herbert R. Halvorson. Biochemistry , 1979, 18 (12), pp 2480â...
0 downloads 0 Views 1MB Size
2480

HA L V 0 R S 0N

B I OC H E M ISTRY

Kirschner, K., & Voigt, B. (1968) Hoppe-Seyler’s Z . Physiol. Chem. 349, 632-644. Klein, W. (1935) Z . Physiol. Chem. 231, 125-148. Knivett, V. A. (1954) Biochem. J . 56, 606-610. Koshland, D. E., Jr. (1952) J. A m . Chem. SOC. 74, 2286-2292. Kresge, A. J., & Yang, Y. C. (1977) J . Org. Chem. 42, 7 57-7 59. Lipmann, F., & Tuttle, L. C. (1947) Arch. Biochem. Biophys. 13, 373-377. Long, J. W., & Ray, W. J., Jr. (1973) Biochemistry 12, 3932-3937. Meunier, J.-C., & Dalziel, K. (1978) Eur. J . Biochem. 82, 483-492. Needham, D. M., & Pillai, P. K. (1937) Biochem. J . 31, 1837-1851. Negelein, E., & Bromel, H. (1939) Biochem. Z. 303, 132-1 14. Negelein, E., & Bromel, H. (1969) in Biological Phosphorylations (Kalchar, H. M., Ed.) pp 102-1 14, Prentice-Hall. Englewood Cliffs, N J . Orsi, B. A,, & Cleland, W. W. (1972) Biochemistry 11, 102-1 09. Porter, W. R., & Trager, W. F. (1977) Biochem. J . 161. 293-302. Pugh, W. (1929) J . Chem. Soc., 1994-2001. Richmond, T. G., Johnson, J. R., Edwards, J. O., & Rieger, P. H . (1977) Aust. J . Chem. 30, 1187-1194.

Sasaki, Y., Lindquist, I., & Sillen, L. G. (1959) J . Inorg. Nucl. Chem. 9, 93-94. Schwarzenbach. G . (1958) J . Inorg. Nucl. Chem. 8, 302-312. Segal, H . L.., & Boyer. P. D. (1953) J . Biol. Chem. 204, 265-28 I . Segel, I. H. (1975) Enzyme Kinetics, Wiley-Interscience, N.Y. Seydoux, F. J., Kelemen, N., Kellershohn, N., & Roucous, C. (1976) Eur. J . Biochem. 64, 481-489. Slocum, D. H . , & Varner, J . E. (1960) J . Biol. Chem. 235, 492-495. Storer, A. C.. & Cornish-Bowden, A. (1974) Biochenz. J . 141, 205-209. Sutton, L. E., Ed. (1958) Tables of Interatomic Distances and Configuration in Molecules and Ions, No. 1 1, The Chemical Society, London. Teipel, J . , & Koshland, D. E., Jr. (1970) Biochim. Biophys. .4cta 198, 183-191. Travers, A,, & Malaprade, L. ( 1926) Bull. Soc. Chim. Fr. 39. 1553-1573. Trentham, D. R. (1971a) Biochem. J . 122, 59-69. Trentham, D. R. (1971b) Biochem. J . 122, 71-77. Velick, S . F.. & Hayes, J. E., Jr. (1953) J . Biol. Chem. 203, 545-562. Warburg, O., & Christian, W. (1939) Biochem. Z . 303,4&68. Winter, E. R. S., Carlton, M., & Brisco, H . V. A. (1940) J . Chem. Soc., 131-138.

Relaxation Kinetics of Glutamate Dehydrogenase Self-Association by Pressure Perturbation? Herbert R. Halvorson

ABSTRACT: The kinetics of self-association for beef liver glutamate dehydrogenase (EC 1.4.1.3) have been measured by using pressure perturbation in both the time domain and the frequency domain by monitoring scattered light intensity. The kinetic behavior is entirely consistent with the random self-association model proposed by Thusius et al. [Thusius, D., Dessen, P., & Jallon, J. M. (1975) J . Mol. Biol. 92, 413-4321. The activation volume AV for association is estimated to be

positive, and it is shown that this provides further corroboration of the molecular mechanism advanced by these same authors. A rapid shift in scattered light intensity is attributed to preferential interaction between the phosphate anion and the protein, proceeding with a positive volume change (2-5 mL/mol of phosphate). A description of the instrument developed for this study is also included.

B e e f liver glutamate dehydrogenase (EC 1.4.1.3) (GDH)] exists in solution as a molecule of 336 000 molecular weight, each molecule consisting of six apparently identical polypeptide chains. These molecules of GDH self-associate to form a concentration-dependent statistical distribution of rodlike aggregates, with a single equilibrium constant describing the addition of a unit to a growing chain. This phenomenon has been studied extensively and is covered in recent reviews

devoted to G D H (Sund et al., 1975; Eisenberg et al., 1976). Although the mode of the interactions has been well characterized as a linear indefinite (or isodesmic) self-association by a variety of equilibrium techniques, equilibrium measurements, of course, leave the question of mechanism unresolved. For G D H self-association, it would be interesting to know if any particular oligomeric species plays a special role in the self-association process. The initial stopped-flow kinetic measurements (Fisher & Bard, 1969) showed uniphasic kinetics, suggesting the existence of but a single kinetic interaction. A more detailed temperature-jump study (Thusius et al., 1975) corroborates Fisher’s observation and points out

‘From the Department of Biochemistry and Molecular Biology, Edsel B. Ford Institute for Medical Research, Detroit, Michigan 48202. Received November 21, 1978. This work was supported by Grant GM 23302 from the National Institutes of Health and in part by an institutional grant to the Henry Ford Hospital from the Ford Foundation. The purchase of the digital signal averager was made possible by an equipment grant from the Ford Motor Company Fund.

0006-2960/79/04 18-2480$01 .OO/O



Abbreviations used: GDH, bovine liver glutamate dehydrogenase (EC 1.4.1.3).

C 1979 American Chemical Society

KINETICS OF SELF-ASSOCIATION FOR GDH

the mechanistic implications of the way in which relaxation times vary with total protein concentration. The self-association equilibrium for G D H is consistent with two quite different kinetic schemes (Thusius, 1977): sequential indefinite self-association, in which growth of a chain proceeds solely by the addition of monomers, and random indefinite selfassociation, in which growth occurs by the combination of any oligomers. The random scheme predicts the observation of the single elementary relaxation. The sequential scheme predicts the occurrence of an infinite relaxation spectrum which, because of degeneracy and relative amplitudes, would be interpreted as a single relaxation. The way in which relaxation rates vary with total protein concentration supports the random indefinite self-association model. In dealing with such a polydisperse system one must make simplifying assumptions, and for the temperature-jump studies just mentioned, the assumptions pertain to the distribution of enthalpy changes among the interactions. Although the assumptions are reasonable, it is not inconceivable that a different distribution of the enthalpy changes and a different kinetic scheme might have led to the same observation. If there truly were heterogeneity in the interactions, it is unlikely that an independent thermodynamic property, volume, would be distributed in the same way as enthalpy. The pressure perturbation study reported here therefore provides strong substantiation for the random model. It also provides additional mechanistic information not revealed by other techniques. The kinetics of protein self-association reactions have been studied previously by pressure-jump light-scattering experiments (Kegeles & Ke, 1975; Davis & Gutfreund, 1976; Tai et al., 1977; Kegeles, 1978). These experiments have entailed monitoring the exponential response to a single rapid compression or decompression of 100 atm or more. Because of the relative insensitivity of most chemical systems to pressure, the resulting perturbation of the equilibrium is comparable to that occurring in a typical temperature jump of 5 OC. For many systems this is a simple and direct way to obtain the desired kinetic information. In other cases, however, it may be desirable to avoid the elevated pressures and large jumps because of concerns about protein stability or a desire to produce only a minute perturbation of a very cooperative transition. Clegg et al. (1 975) have used repetitive, small (less than 10 atm) pressure perturbations, with signal averaging techniques, to study order-disorder transitions in lipid vesicles and proteins. I have built a similar instrument and applied it to the self-association of GDH. Although the need for a small perturbation is not great in this particular case, the capability is important for study of other, more cooperative polymerizations such as those which are displayed by flagellin, myosin, actin, tubulin, or deoxyhemoglobin S. Experimental Procedure Materials. G D H was obtained from Sigma (type IV). After dialysis, solutions were clarified by centrifugation and used without further purification. Other chemicals were standard reagent grade. For all experiments reported here the buffer was 0.2 M potassium phosphate and 1 m M EDTA, p H 7.2. Concentration of G D H was determined spectrophotometrically WrnL = 0.97 (Olson & Anfinsen, 1952)]. Light scattering was monitored a t 400 nm. Pressure Perturbation Instrument. The design of the instrument parallels that of the instrument developed by Clegg & Maxfield (1976) in its essential features. Figure 1 is a block diagram of the configuration for measurements in the frequency domain (dispersion). A pressure wave sinusoidal in

VOL.

18,

NO. 12, 1979

2481

HP-203A-001 Variable Phase Function Generator

Phase-Sensitive Detector Integrator

DVM ( Q )

----------

I

I

Phase-Sensitive Detector Integrator

I

offset or reference PMT FIGURE 1 : Block diagram of pressure relaxation instrument. Monochromatic light isolated from a Xe arc impinges on a short cylindrical cell. Alternate positions of the detector allow the monitoring of transmitted light or light emitted at 90'. Pressure is varied by driving a piezoelectric stack with the output of a programmable power supply. A variable-phase variable-frequency function generator synchronizes the pressure perturbation with the phase-sensitive detector/integrator. For time-domain experiments (not shown) the power supply is programmed with a square wave and the detector/integrator is replaced with a digital signal averager.

time is generated by driving a stack of piezoelectric ceramics (LTZ-1, Transducer Products) with the high-voltage output of an operational power supply (OPS-2000, Kepco). The power supply in turn is driven by a variable-frequency, variable-phase function generator (HP-203A-001, HewlettPackard) which provides all the necessary timing information for the circuitry. The sample, contained in a short cylindrical optical cell (synthetic sapphire, Adolf Meller Co.) is illuminated with monochromatic light (H-20 holographic grating 'monochromator, Jobin-Yvon) from a 150 Xe arc (Osramj in an Oriel housing, powered by a constant-current power supply (XL 150, O.L.I.S.). Light scattered a t 90° is detected with an EM1 9558QB end-on photomultiplier tube in a PPI 62/3A27 combination photomultiplier tube housing/photometer amplifier. Conversion to detection of fluorescence (emission filter) or transmittance (in-line observation) is simple. A reference photomultiplier tube samples the incident light, giving the capability of removing most of the lamp fluctuations. The difference between the two photomultiplier signals is taken with a Keithley 301 operational amplifier. A calibrated semiconductor bridge pressure transducer (XTM- 1- 190-500, Kulite Semiconductor Products, Inc.) monitors the pressure within the cell. When the instrument is operated in the frequency domain, it employs fairly standard techniques of phase-sensitive detection and amplification in a homemade dual-channel device (Bentz et al., 1974; Mathis & Buck, 1976). Instead of the

2482

B I O C H E 2.1 I S T R Y

usual active filter, this circuit employs a gated integrator in the final stage. This approach obviates the attenuation of the signal to root mean square level and the difficulty of filtering out low-frequency ripple-two major problems with traditional lock-in techniques under these conditions (low signal and low frequency, respectively). For kinetic measurements in the time domain, the piezoelectric stack is driven with a square wave, producing repetitive transient signals. The output of the Keithley 301 is then supplied to a digital signal averager (Tracor-Northern 507A), where successive transients are accumulated until a satisfactory ratio of signal to noise is achieved. The oscilloscope display is then photographed. The enhancement of the ratio of signal to noise differs for the time domain and frequency domain techniques and is treated separately and more completely under Results. The extent to which random noise is reduced increases as the square root of the number of cycles. Sudden jumps in lamp intensity resulting from arc wander can be disastrous, but the differential measurement greatly reduces the influence of more minor (and more common) fluctuations. Smooth variations in lamp intensity (drift) will influence the total amplitude only, with no effect on rates or the fractional amplitudes of part processes. The stainless steel bomb, which contains the piezoelectric stack, the optical cell, and the pressure sensor, was not thermostated in these experiments, although that capability now exists. Instead, its large thermal mass [7 kg, with a specific heat of 0.1 cal/(g deg)] was relied on to maintain the experimental temperature at room temperature (20 “C). Since the pressure-jump experiment is adiabatic rather than isothermal, it is appropriate to consider the validity of this assumption. Adiabatic heating of the sample is described by ( a T / a ~=) ~aT/pcP which for water at 20 OC becomes 1.5 X deg/atm. This effect is insufficient to sensibly perturb the chemical system and is canceled over each complete cycle. Cumulative electrical heating is diminished by the small power factor for the piezoelectric stack, which behaves like an almost pure capacitance. At 2 kV the dc current is less than IO-’ A, which for a I-Hz square wave (50% duty cycle) is equivalent to about cal/cycle. The net resistive heating over 2000 cycles would be less than OC. Theory. The principles of relaxation kinetics are familiar: by effecting a small perturbation of a chemical system at equilibrium, the rate a t which the system relaxes to the new equilibrium conditions can be analyzed by using greatly simplified kinetic expressions (Eigen & de Maeyer, 1963, 1973; Bernasconi, 1976). The observed intensity can be expressed as Z, = HcM, where Z, is the observed (static) intensity of scattered light, c is the mass concentration of macromolecule, M , is the weight-average molecular weight of the solute, and H i s the familiar “instrument constant” H =2~~n~~(dn/dc)~/(N~h~r~)

Since only small departures from Io are measured, it is legitimate to ignore the virial terms. The amplitude resulting from a small perturbation dp is given by i = Zo(d In Zo/dp)Gp

Measurements can be performed in the time domain (transient techniques) or the frequency domain (stationary techniques). Transient techniques (“jump” experiments) are characterized by an effectively instantaneous change in the perturbing

H A L v 0 K s0 N

FIGURE 2: Phasor representation of frequency-domain pressure perturbation. The real axis is defined by the pressure perturbation. Physical response iphys lies on the real axis, whereas chemical response ichem is delayed by phase angle 4. The photomultiplier tube responds to islgnal,the vector sum of iphys and ichem.Phase-sensitive detection resolves islw into its two orthogonal components iRand iQ The dashed line is the locus of islgnalas a function of frequency.

variable, followed by an exponential decay in the observable to the new equilibrium. The stationary techniques (absorption or dispersion experiments) typically use a sinusoidally modulated perturbing variable, so the position of the equilibrium becomes a function of time. As the perturbation is swept or scanned from low to higher frequency, the observable signal is attenuated and shifted in phase. The relations are conveniently visualized in a phasor representation. Figure 2 illustrates a hypothetical pressure-modulation experiment on a self-associating system with a positive volume of association, by using scattered light as the observable. Each vector represents the amplitude and phase angle of a particular harmonic oscillation of angular frequency w, with the real axis being defined by the perturbation. The physical response to the perturbation is effectively instantaneous and lies on the real axis. The signal depicted iphyswould pertain to an increase over the dc level due to the finite compressibility of the solvent. Changes in solvation producing changes in the refractive index increment would lead to positive or negative amplitudes for iphys.The chemical response lags behind the perturbation by phase angle J#C = tan-’ ( U T ) , the finite rate precluding instantaneous response. This “chemical compressibility” leads to a negative amplitude in the signal &hern because a positive volume change means that a n increase in pressure drives the system to lower M,, (less scattered intensity). The photomultiplier tube sees islgnal,the vector sum of iphysand ichem. Depending on the relative magnitude of iphys,the apparent phase angle can deviate markedly from 4. Resolving islgnalinto its two orthogonal components iq and iR with phase-sensitive detection avoids this problem since iq contains no contribution from iphys.The dashed line indicates the locus of islgnalas a function of frequency (Cole-Cole plot). When iq and i R are displayed as functions of frequency (Figure 31, conventional dispersion curves result. Analysis of experiments in the time domain generally follows the treatment developed for single transients. Particularly when dealing with slow relaxations, it can be advantageous to truncate the decay before it reaches the equilibrium value in order to collect more transients a t the expense of decreased amplitude. The complete expression for the signal observed from a single normal relaxation* is

V O L . 1 8 , N O . 1 2 , 1979

KINETICS O F SELF-ASSOCIATION F O R G D H 0

I

R

I 0.5

0.1

5

1

10

WlWo

Dispersion curves for i and. . i The real component is given by iR = [B - (1 + w2$)-'rA%/RT]6p and the quadrature or imaginary component by iQ = [ w r ( l + w'rz)-'rAv/Rr]6p. FIGURE 3:

5L

llftrefl Response of phase-sensitivedetector and gated integrator to inputs of differingratios to the operating frequency. Inset: power sparum of environmental noise showing 1/J interference, and "white" noise. FIGURE 5:

Relaxation of GDH (5.8 mg/mL) in the time domain (2048 transients). The optical signal is inverted by the differenceamplifiier, so the initial pressure drop (6 atm) produces an increase in the scattered intensity. The half-period of the square wave is 244 ms (f = 2.05 Hz). Amplitudes of the two traces are arbitrary. FIGURE 6:

' ?

1

0.5

0.05 0.1 0.2

0.5

0.1 0.05 0.02 0.01 1 0.005Lrtsec,

0.2 1

2

f

(HA

5

10

20

50

FIGURE4 Relaxation of GDH (10 mg/mL) in the frequency domain. Points are the means (iSE)of IO measurements. Solid lines are theoretical curves for a process with an amplitude of 24 mV and a relaxation time of 32 ms. See text for a discussion of data at 20 Hz.

where T is the chemical relaxation time, T~ is the time constant for the exponential change in pressure, and T i s the duration of the perturbation (half the period of a square wave). These corrections for finite rise time and truncated decay were not required in the experiments reported here. Results Frequency Domain. A representative dispersion curve is shown in Figure 4. The theoretical curves were obtained by first defining the amplitude with a Cole-Cole type plot and then assigning the maximum in iQ t o . l / r . This yields a relaxation time of 32 ms. The fit appears adequate, although least-squares treatment was not employed. The data at 20 Hz, not used in estimating 7 , illustrate a point to be derived from Figure 5, which represents the relative discrimination of the amplifier/integrator. The signal entering this device contains noise components of many frequencies oi besides the com-

'This mrrsfsa minor typographical error in eq 15 of Clegg Br Mafield

(1 976).

2483

ponent a t the operating frequency 0. The output voltage is derived by full-wave rectification (at frequency a), followed by integration over one complete cycle at the same frequency. The contribution of noise at frequency oi to the output voltage is expressed as

N, =

In'* sin 0

( w i t ) dt -

Lysin ( w i t ) dt

and it is the absolute value of this function, normalized to unity at a, which is depicted in Figure 5. Not all noise frequencies are equally attenuated. When operating a t 20 Hz, line frequency (60 Hz) is the first odd harmonic and it is not effectively removed. Accordingly, the values of iQ and iR at this frequency are strongly influenced by the phase relation between the pressure perturbation and the line voltage. The greatly increased uncertainties in the mean values at this frequency arise from a ratio of interference frequency to operating frequency which is not exactly 3. This produces beating. The data a t 17.5 Hz (60117.5 = 3.4) illustrate the ability of the analyzer to discriminate against a strong noise component under more favorable conditions. Although not shown here, even more striking effects are seen in the vicinity of 8 Hz,since the dominant noise frequency is 120 Hz (12018 = 15). The presence of l l j o r flicker noise and the finite low-frequency

2484

.

HALVORSOU

B I OC H E M I STR Y

’01

C8

-

06

3

0 (i.

I

L

04

02

1000

-

aoo

/=-

-

_______________

600

-

w 0

A

ci)

c 2 400

-

200

I

2

4

6

8

1

2

0

C T (mgirnl)

FIGURE 7 : Fraction of the total change in optical signal due to the relaxation amplitude vs. total protein concentration. Points are experimental results, and lines are theoretical as described in the text.

lobe for the analyzer response combine to preclude measurements much below 0.05 Hz. This is a general problem with techniques in the frequency domain, particularly when the ratio of signal to noise is low. Time Domain. The results of a typical transient (time domain) experiment are shown in Figure 6. Although white (random) noise is suppressed as the square root of the number of repetitions, the relative contribution of interference noise will be enhanced if its frequency is a harmonic of the frequency of the square wave. These harmonics become closely spaced a t low operating frequencies, leading to the appearance of “ringing” in the signal. Continued averaging of more transients cannot eliminate this from the record, so operating frequencies must be chosen carefully. Another feature of the pressure-jump relaxation pattern is the presence of a fast process (following the time course of the perturbation, rp = 2 ms) with an amplitude of the same sign as the relaxation proper. Although this phenomenon was not noted in temperature-jump studies on this system (Thusius et a]., 1975), similar effects have been seen in pressure-jump experiments with a-hemocyanin and attributed to changes in solvation involving larger volume changes than enthalpy changes (Tai et al., 1977). Preferential interaction of GDH with phosphate has been noted to increase the refractive index increment (Gauper et al., 1974). If this is the underlying cause of the fast process, then the preferential interaction between the two components proceeds with a net increase in volume. Such behavior would be consistent with removing the phosphate anion from “normal” aqueous solution to a region where less electrostriction would be obtained, i.e., in the immediate proximity of the protein molecule. Relaxation curves were analyzed by plotting semilogarithmically points taken from the photographs and drawing a line through the points at longer times. “Curve peeling” was used to verify that the fast process occurred at the same rate as the perturbation. The fraction of the total amplitude caused by the slow phase was derived from the same plot. This information is presented in Figure 7 .

6

4

CT

8

1 0 1 2

(mgml)

Reciprocal of the square of the relaxation time vs. total protein concentration. Experimental points are from measurements or frequency domain ( 0 ) .Where relaxation in the time domain (0) times differ at the same concentration, the slower process always corresponds to the higher equilibrium pressure. F I G ~ ~ R8:E

Random indefinite self-association (Thusius et al., 1975; Thusius, 1977) can be regarded formally as a dimerization reaction between sites available for interaction, S, to form a bond or tight interaction, B,

----

2s & B kd

where the numerical factor arises from statistical considerations. As a consequence of this formal dimerization, the familiar transformation can be applied to test the concentration dependence of the experimentally observed rates and to derive the rate constants in the absence of ancillary information. The relaxation data are presented this way in Figure 8, which shows the equivalence of the time domain and frequency domain techniques. From the slope and intercept, I determine k, = 1.5 X lo6 M-I s-’ and kd = 5 s-l, equivalent to values determined earlier (Thusius et al., 1975). Since each pressure-jump experiment gives two relaxations (at different pressures) and the rate at the higher pressure is consistently slower, the activation volume for self-association can be estimated from the pressure dependency of the rates. Differentiation of the expression for the relaxation time yields 1 A In ( 1 / ~ ) AVa* = - [ 1 + ( ~ k , j ) * ] ~ l -V ‘RT 2 AP Under the conditions of these experiments the thermodynamic volume change for association is +24 mL/mol (Heremans, 1975) and kd is 5 s-]. The result, AV,’ = +350 f 150 mL/mol, presents immediate implications as to the nature of the transition state and the mechanism for the self-association process. Discussion The results obtained here on the kinetics of G D H selfassociation provide substantial corroboration for the random

VOL.

K I N E T I C S O F S E L F - A S S O C I A T I O N FOR G D H

indefinite associate model of Thusius et al. (1975). If the interactions in this system were truly heterogeneous, it would be quite unlikely that both AVand AH would be distributed so as to give an apparent single exponential decay (Figure 6) through degeneracy. The concentration dependence of the relaxation times shown in Figure 8 argues strongly against any special role for the monomer in the kinetic mechanism. Further insight into this system is provided by a closer consideration of the results in Figure 7 and 8. Since the self-association occurs more slowly than a diffusion-controlled process yet the subunits must encounter one another in some diffusion-limited manner, it is helpful to consider a slightly more detailed kinetic scheme 2s

S2

18,

NO. 12,

1979

2485

becomes indistinguishable from 1 . This is not in accord with the results shown in Figure 7. Compressibility of the solvent can be immediately rejected as an explanation because it would produce a jump opposite in sign to the relaxation amplitude. A remaining plausible explanation is a decrease of refractive index increment with increasing pressure. Since scattered intensity varies as the square of the refractive index increment, the amplitude expected is 61 = -2I~Dbp where D = -a In (dn/dc)/ap. The relaxation proper will have an amplitude (Thusius et al., 1975) of

B

where, as before, S represents a site available for interaction and B represents a bond or tight interaction. S2 represents an encounter pair produced by the collision of two hydrated sites. Step 1 proceeds at a diffusion-limited rate, while the unimolecular conversion in step 2 is much slower, akin to the mechanism for ligand-metal complex formation (Eigen & Tamm, 1962). Since S and S2 represent heterogeneous molecular populations, one cannot speak meaningfully of a single diffusion-controlled rate constant for step 1 , but it is possible to estimate that the relaxation time for this step should not exceed 100 ps, which is significantly faster than the perturbation. Such a mechanism is consistent with all that is known about G D H self-association. The fast relaxation process summarized in Figure 7 does not arise from step 1 of this mechanism, however, as can be shown by analyzing the amplitudes. Denoting by Adl and AK the molar changes in light intensity and volume associated with each part reaction, the fraction of the total amplitude which occurs in the slow phase can be expressed (Schimmel, 197 1; Thusius, 1972) as

f = [(A41 - A42) x

+

+

(AVl - AV2) A42AV24KIs - b$lAV14Kls/(l 4K1s)l/[(A41- A42)(Av1 - AVA + A42AV24Kls + A4lAVI/K21 where s = (v‘l + 4KlK2C, - 1)/(KIK2) is the concentration of free sites. The A4’s can be expressed as

A41 = 2MI2[2H(S2) - H(S)] A42 = 4M12[H(B) - H(S2)I where allowance has been made for the “instrument constant” H (or, by implication, dnldc) to be a function of the degree of association. Such a formulation is at odds with the well established idea that the dependence of light scattering on molecular size stems from interference effects in the solute particle and accordingly should be insensitive to the minute change implied for step 2. This implies that

1 AV K --Cl f=

DRTM,

This expression is plotted in Figure 7 for different limiting values o f f , and from the solid line it appears that an appropriate estimate is 0.5 or D = 2.5 X Preferential interaction of G D H with phosphate has been noted previously (Gauper et al., 1974) through the effect of phosphate concentration on the refractive index increment of the protein. Analyzing their data through the procedures described by Casassa & Eisenberg (1964) leads to an estimate of 180 f 5 mol of phosphate per mol of G D H in 0.5 M phosphate buffer, pH 6.5. (This represents 30 mol of phosphate per polypeptide chain, in reasonable accord with 26 mol of NaCl per mol of BSA under similar conditions.) Using the empirical formulas of Teller (1976) for the accessible surfaces of proteins, the fraction of the surface which is occupied by phosphate (0.03) does not differ materially from the estimated volume fraction of phosphate in the solution (0.05). Indirect support for the idea that these are nonspecific interactions comes from the relative insensitivity of the self-association to the phosphate concentration, d In K/d In (phosphate) = -0.92 at 10 OC, pH 6.5 (Gauper et al., 1974), suggesting that only a negligible fraction of the bound phosphate is linked to the self-association, perhaps involving specific interactions. (The kinetic uncoupling of the two processes is established by the experiments reported here.) It is then reasonable to suppose that these are weak, nonspecific interactions involving one R T (0.6 kcal) of excess free energy for each phosphate (Schellman, 1978). In the absence of a specific binding isotherm, the thermodynamic measure of binding reduces to am3 - = k2(excess)/RT am2

A41 = 4MI2H A42 = 0 leading to

The nature of the two steps implies that AVl

The fraction of the amplitude occurring in the slow phase is then